Océ Technologies

Océ Technologies, with 22,000 employees world-wide and annual sales of EUR 3.2 billion, is a multinational technology company that helps organisations manage and control their document flows, and provides a range of professional hardware, software and business services for added-value document handling.

Océ products and services are available in some 80 countries, and the company supports global customers by providing user manuals in up to seventeen languages and technical service manuals in five languages.

Introducing Language Technologies
Océ moved into automated translation due to increasing translation volumes and the high cost of manual translation. Lou Cremers, then a member of Océ's natural language processing group, built a prototype system that convinced his management of the benefits of MT.

As a result, Lou Cremers started a translation services facility and decided to invest in a commercial MT system from Logos, which was chosen because of its relatively good performance and flexibility. As he points out: "For partly historical reasons, we started at the backend of the translation process, with the introduction of MT. It is not the best place to start, but it did not require major changes in existing workflows". Soon afterwards, the system was enhanced though the addition of Trados translation memory (TM) software.

Although automated translation through MT and TM led to a substantial reduction in translation time and cost, there was still room for improvement. The company was dealing with an increasing number of types of documents to be translated, ranging from software to online help and Lotus Notes databases. Each document format required specific conversions to enable the translation of each source document through a common workflow.

Many of the problems encountered were the direct consequences of typical second language mistakes, of differences in style, and of the inconsistent use of terminology. Since documents were not generally available online, there was little reuse of documentation, except for updates. At this point it became clear that in order to further increase quality, speed, and cost efficiency, Océ had to re-examine the documentation creation process, and introduce changes going far beyond the translation phase.

Central in the new approach is a multilingual XML document-management database, which typically contains modular information elements. These information elements are subsequently combined to form specific publications, whether as HTML, a help file, or a user manual. Changing the workflow is an ongoing process, and the implementation of XML is expected to be complete in about one year from now.

In order to implement the new approach, it was essential to develop a completely new way of working. To achieve this, two modules were introduced before the actual XML implementation to ensure optimal standardisation and control of written texts: Text & Structure and Controlled English.

Text & Structure is an Océ-internal course introducing the general principles of how to present information and how to group, label, and "chunk" information. Controlled English (CE) describes how to formulate the information elements and in that sense is complementary to Text & Structure. It follows the AECMA Simplified English rules, but has been adapted to the Océ environment in a less strict form.

Terminology
In order to maintain terminological consistency and correctness in both the authoring process and the translation process, a lot of effort was put into the development of a database to make technical dictionaries, and more recently into a bilingual terminology mining tool, available on the Océ intranet. "The importance of a terminology database is often underestimated. It is the basis of any quality or efficiency improvement in documentation and translation", says Lou Cremers.

The introduction of Controlled English facilitated the standardisation of technical terminology, which led to the development of a custom-built multilingual database that allowed users to not only browse terminology, but also to access all available translation memories in up to 17 languages. Once source terminology in English has been established, unique and consistent translations are needed in a number of languages. In order to achieve this, the bilingual terminology miner searches the translation memory for relevant terminology and the statistically most appropriate translations. A native speaker then selects the correct translation, and can also add complementary information like class and gender. This enables a fast build-up of multilingual corporate terminology, which is also stored in the multilingual database.

Improvement Measurement and Workflow Automation
The translation services unit has also developed a tool that measures the efficiency of the automation process. The tool kept track of the traditional translation cost and the actual cost of a translation project after deploying MT and TM tools. It has now evolved into a workflow tool that not only tracks costs and savings, but also automates most of the workflow.

Through an intranet web application, project co-ordinators create the project data infrastructure and fill in a translation request form, together with all the information required to start the translation cycle. First, a custom built tool analyses the source documentation against all translation memories and identifies the most relevant specific memories. These results are then analysed to produce the metrics for the project.

The workflow tool reads the metrics and calculates the cost and duration of the project. The actual translation is outsourced, and a translator package is generated for the project that contains source, memories, log files and an automatically created readme file that contains the instructions and other details. With the push of a button, the package is put on Océ's FTP server, and the translator allocated to the job is notified. The tool also makes it possible to keep track of each project's status, to store the resulting translation memories, and to archive the project automatically on completion.

Benefits
Thanks to the new documentation process, the quality of documentation has improved considerably. The documentation is clearer, more concise, and more consistent in the use of terminology. A side-effect of using the CE tool is a reduction of superfluous information - the "need-to-know" as opposed to the "nice-to-know". Instead of thinking in terms of the final publication, authors now think in terms of information modules. Although CE constrains creativity to some extent, most authors recognise its usefulness.

Reuse of existing text has improved since the online database contains publication-independent generic information modules. Because the source material meets a higher quality standard, the MT system produces much better draft translation for FIGS languages (French, Italian, German, and Spanish), and translators need less time to post-edit MT drafts.

Finally, the authors are freed from DTP work, and can fully concentrate on content. As Lou Cremers points out "The success of MT in a corporate environment partly depends on the simultaneous introduction of a new mind-set. MT should be regarded as enabling technology which speeds up routine work, leaving the translator to apply special domain and language skills".

This success story is based on an interview with Lou Cremers, Translation Services Manager at Océ Technologies. We would like to thank him very much for his help and co-operation.

Better, Cheaper, Faster Translation at Océ Technologies